{"id":169834,"date":"2019-10-25T00:10:24","date_gmt":"2019-10-24T16:10:24","guid":{"rendered":"https:\/\/www.geek.com\/featured\/-1808445\/"},"modified":"2019-10-25T00:10:24","modified_gmt":"2019-10-24T16:10:24","slug":"ai-better-than-doctors-at-detecting-brain-hemorrhages","status":"publish","type":"post","link":"https:\/\/itteacheritfreelance.hk\/test\/wordpress\/2019\/10\/25\/ai-better-than-doctors-at-detecting-brain-hemorrhages\/","title":{"rendered":"AI Better Than Doctors at Detecting Brain Hemorrhages"},"content":{"rendered":"
\n\t\t\t\t\t<\/p>\n
A deep learning algorithm recognizes abnormal CT scans of the head in neurological emergencies in 1 second. (Photo Credit: University of California, San Francisco)<\/div>\n<\/figure>\n

This algorithm is better at finding tiny brain hemorrhages than human experts.<\/span><\/p>\n

The technology<\/a>, developed by scientists at UC San Francisco and UC Berkeley, could one day help doctors treat patients with traumatic brain injuries, strokes, and aneurysms.<\/span><\/p>\n

Radiologists look at thousands of head scans a day, searching for itty-bitty abnormalities that can signal life-threatening emergencies.<\/span><\/p>\n

That’s a lot of pressure.<\/span><\/p>\n

Enter an algorithm.<\/span><\/p>\n

It’s a lot more efficient (and potentially more accurate) for artificial intelligence<\/a> to scroll through stacks of images, flagging those with significant abnormalities and passing them on to radiologists for further examination.<\/span><\/p>\n

“We wanted something that was practical, and for this technology to be useful clinically, the accuracy level needs to be close to perfect,” study co-author Esther Yuh, associate professor of radiology at UCSF, said in a statement.<\/span><\/p>\n

“The performance bar is high for this application, due to the potential consequences of a missed abnormality,” she continued. “And people won’t tolerate less than human performance or accuracy.”<\/span><\/p>\n

Their final product, detailed in a paper published this week by the journal <\/span>Proceedings of the National Academy of Sciences (PNAS)<\/span><\/i><\/a>, took just one second to determine whether an entire head scan contained any signs of hemorrhage.<\/span><\/p>\n

It even found some small abnormalities that experts missed.<\/span><\/p>\n

The algorithm also noted their location within the brain<\/a> and class<\/a>ified them according to subtype\u2014and did it all with “an acceptable level” of false positives.<\/span><\/p>\n

Now it’s just showing off.<\/span><\/p>\n

To make this clinically useful, though, the technology must achieve 100 percent accuracy.<\/span><\/p>\n

“If a computer is pointing out a lot of false positives, it will slow the radiologist down, and may lead to more errors,” Yuh explained.<\/span><\/p>\n

The key, according to Jitendra Malik, professor of electrical engineering and computer sciences at Berkeley, was choosing which data to feed into the model.<\/span><\/p>\n

Using a fully convolutional neural network (FCN), researchers trained their algorithm on a relatively small number of images\u20144,396 CT exams, each image was packed with information.<\/span><\/p>\n

The richness of this data, according to the team, helps prevent “noise” to create an extremely accurate system.<\/span><\/p>\n

“We took the approach of marking out every abnormality\u2014what’s why we had much, much better data,” Malik, a co-author of the study, said. “Then we made the best use possible of that data. That’s how we achieved success.”<\/span><\/p>\n

Moving forward, researchers are applying the algorithm to CT scans from trauma centers across the country.<\/span><\/p>\n

More on Geek.com:<\/span><\/i><\/p>\n